关键词: 超细全尾砂, 固结粉, 充填体强度, 优化配比, 极差分析, 神经网络

Abstract: In order to reduce the costs of filling mining in Sishanling Iron Mine， the ultra-fine solid tailings of mines， with slag and desulfurized gypsum in Benxi area as raw materials， were used to study the low-cost consolidated powder filling cementing materials required by the filling mining method. Firstly， the orthogonal tests with 3 factors and 3 levels of salt-base and alkali-base activator gelling materials were designed. The optimized formula of the solidified powder activator for 3 d strength of the filling body was 9% cement clinker， 4% desulfurized gypsum and 1% industrial mirabilite. The range analysis is made to obtain the weights of the influencing factors of the filling bodies 7 d and 14 d， ranking from high to low: cement clinker， industrial mirabilite， and desulfurized gypsum. Then， BP neural network model was established to predict the strength of cement with different proportions of the initiator. The maximum relative error of the model was 4.83%， which satisfies the training accuracy. Then， the quadratic polynomial fitting， the optimized formula for the ultra-fine full tailings consolidated powder filling gelling material is obtained. The optimized formula of consolidated powder filling body for 7 d is 5.88% of cement clinker， 9.31% of desulfurized gypsum， 0% of mirabilite， realizing 2.40 MPa of fillingbody strength； The optimized formula of filling body for 14 d is 5.99% of cement clinker， 9.63% of desulfurized gypsum， 0.40% of mirabilite， realizing the strength 3.12 MPa. The verification test results show that the strength of the cemented cement filling body for 14d is 1.48 times higher than that of the cemented powder， and the cost of the consolidated powder cementing material is lowered by 23%.

Key words: Superfine whole tailings, Consolidated powder, Backfill strength, Optimized ratio, Range analysis, Neural network